You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Tomek Rękawek (JIRA)" <ji...@apache.org> on 2017/01/28 09:57:25 UTC
[jira] [Commented] (OAK-5222) Optimize the multiplexing node store
[ https://issues.apache.org/jira/browse/OAK-5222?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15844007#comment-15844007 ]
Tomek Rękawek commented on OAK-5222:
------------------------------------
OAK-5494 allowed to disable mounting path fragments. If the path fragment support is indeed disabled, we can switch to "native" node state/builder objects as soon as there are no mounts placed under the current path.
Test results:
{noformat}
java -jar oak-run.jar benchmark ReadDeepTreeTest Oak-Multiplexing Oak-Segment-Tar --pathsPerMount=0 --mounts=0
# ReadDeepTreeTest C min 10% 50% 90% max N
Oak-Segment-Tar 1 14 14 15 17 28 3821
Oak-Multiplexing 1 13 14 15 17 29 3904
{noformat}
So, for the 0-mount setup the MultiplexingNodeStore overhead is close to 0.
> Optimize the multiplexing node store
> ------------------------------------
>
> Key: OAK-5222
> URL: https://issues.apache.org/jira/browse/OAK-5222
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: core
> Reporter: Tomek Rękawek
> Assignee: Tomek Rękawek
> Fix For: 1.8
>
> Attachments: OAK-5222-1.patch, results1.png
>
>
> I added support for the multiplexing node store for the oak-run benchmark fixtures. It seems that the performance in ReadDeepTreeTest is linearly dependent on the mount counts:
> {noformat}
> Mounts N
> 0 729
> 1 402
> 2 287
> 3 209
> 4 188
> 5 154
> 6 133
> 7 126
> 8 104
> 9 100
> 10 85
> 20 46
> {noformat}
> {{N}} - throughput in 60 seconds. Following command was used:
> {noformat}
> $ java -jar target/oak-run-1.6-SNAPSHOT.jar benchmark ReadDeepTreeTest Oak-Multiplexing --mounts=10
> {noformat}
> It should be possible to improve this, so we won't have a noticeable performance decrease for 1000 mounts.
> //cc: [~rombert]
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)